Features based classification of hard exudates in retinal images
The people suffering from diabetic retinopathy is increasing continuously over the time. There is need for huge number of experienced ophthalmologists to identify and treat the diabetic people at early stage of retinopathy. There are many features of retina those indicate progress of diabetic retinopathy such as blood vessels, optic disc (OD) which are bright objects that changes when eye is affected by disease and it can be used as main features for detecting diabetics' presence. In our work a new method for identifying and separating out the optic disc from fundus RGB image using wavelet transform and windowing (partitioning) the image is performed. Then energy and standard deviation for each window is computed to segment out optic disc. The proposed method is evaluated by considering DRIVE and STARE datasets. It is observed that the method yields 95% of detection accuracy.